[geos-devel] Test OverlayNG for MVT tiling

Martin Davis mtnclimb at gmail.com
Mon Aug 3 11:48:02 PDT 2020


Great that you're testing this.

Yes, currently the precision is not exposed via the C API (well, I think it
is technically, by setting the geometry precision model, but this involves
an expensive (and unnecessary) precision reduction step, so that's useless
for performance.

My hope is that the C API will expose overlay operations which take an
explicit precision (scale factor = 1 / grid size). This will invoke
snap-rounding, which should provide fully-robust rounded geometry - ideal
for clipping to low-resolution tiles - I hope!


On Mon, Aug 3, 2020 at 10:19 AM <rmrodriguez at carto.com> wrote:

> I've been doing some tests today and opened a couple of tickets
> related to issues with intersections.
>
> I'm not sure how interesting they are right now as I haven't checked
> for validity, but I've run the default tile generation performance
> test (full tiler) to compare the performance between the new and old
> overlay generations (attached doc).
>
> The MVT tests are using the GEOS backend, which is still much slower
> than wagyu (5-10x in the slow cases, which are huge polygons in low
> level zooms) but there are some really nice improvements there (see
> `[MVT] Canada [...]` tests). It is possible that those cases are slow
> because they need to call makeValid because the call to Intersection
> fails or because new points are added outside the MVT precision, but
> the only way to check is to add logging and check specific tiles. In
> any case, I've only used the GEOS C API, which AFAIK doesn't yet
> provide control over the precision.
>
> There are other tests that have slowed down a bit but I haven't looked
> into any of them at this point.
>
> --
> Raúl Marín Rodríguez
> carto.com
> _______________________________________________
> geos-devel mailing list
> geos-devel at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/geos-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/geos-devel/attachments/20200803/4f0f02c3/attachment.html>


More information about the geos-devel mailing list